at the moment, the coda-api will be selected at compile time. Because of
this we need different (Debian)-packages for different kernels.
I have got an idea to change this:
The differences in the coda api, as far as they concern davfs2, seem to
be only the different structures of inputArgs and outputArgs.
We could define for every coda api separately:
- the structures for inputArgs and outputArgs
- a set of get_* and set_* function for every field we need
At runtime we evaluate the coda version and set pointers to the
appropriate set of get_* and set_* functions.
To read or manipulate the inputArgs or outputArgs we use this functions.
This would decrease performance slightly but I think not too much.
What do you think about? Might it work? Is it worth the effort? Some
ideas for a different aproach?